Chemoecology 17: 157–161 (2007)
0937-7409/07/030157-5
' Birkhäuser Verlag, Basel, 2007
DOI 10.1007/s00049-007-0373-0
CHEMOECOLOGY
The paradox of risk assessment: comparing responses of fathead
minnows to capture-released and diet-released alarm cues from
two different predators
Maud C. O. Ferrari, Myles R. Brown, Michael S. Pollock and Douglas P. Chivers
Department of Biology, University of Saskatchewan, Saskatoon, SK, S7N 5E2, Canada
Summary. Many aquatic prey are known to use chemical
alarm cues to assess their risk of predation. In fishes, such
alarm cues can be released either through damage of the epi-
dermis during a predatory attack (capture-released) or
through release from the predator feces (diet-released). In
our study, we compared the importance of capture- versus
diet-released alarm cues in risk assessment by fathead min-
nows (Pimephales promelas) that were naïve to fish preda-
tors. We utilized two different fish predators: a specialized
piscivore, the northern pike (Esox lucius) and a generalist
predator, the brook trout (Salvelinus fontinalis). Handling
time of pike consuming minnows was much shorter than for
trout consuming minnows, likely resulting in less epidermal
damage to the minnows during attacks by pike. In accor-
dance with this, minnows showed a less intense antipredator
response to capture-released cues from pike than capture-
released cues from trout. This represents a paradox in risk
assessment for the minnows as they respond to the special-
ized piscivore, the more dangerous predator, with a less
intense antipredator response. In contrast, the minnows
showed a stronger antipredator response to the specialized
piscivore than to the generalist when given diet cues. This
work highlights the need for researchers to carefully con-
sider the nature of the information available to prey in risk
assessment.
Key words. Chemical alarm cues – predator odour – diet
cues – risk assessment – fathead minnow (Pimephales
promelas, Family Cyprinidae, Order Cypriniformes, Class
Actinopterygii)
Introduction
Due to the unforgiving nature of predation, prey animals are
under intense selection to detect and avoid predators (Lima
& Dill 1990, Wisenden & Chivers 2005). Because respond-
ing to predators is costly, animals displaying adaptive
responses, i.e. optimizing the trade-off between antipredator
behaviour and foraging or reproduction, should be at a
selective advantage (Helfman 1989, Lima & Bednekoff
1999). A prerequisite for effective and adaptive responses
against predators is that prey possess accurate information
regarding the level of threat posed by the predators. For
many aquatic species, chemicals present in the water repre-
sent an important source of information regarding foraging,
reproduction and predation (Chivers & Smith 1998).
Aquatic animals can gather information regarding
predators using chemicals released by the predator, i.e.,
kairomones or predator odours (Kats & Dill 1998). For
example, predator odours have been shown to enable prey
fishes like fathead minnows (Pimephales promelas) to
determine relative size (Kusch et al. 2004), proximity and
density (Ferrari et al. 2006) of predatory northern pike
(Esox lucius). Many fishes, including fathead minnows, do
not have an innate recognition of predators, that is, individ-
uals have to learn to recognize potential predators as a threat
(reviewed by Brown 2003).
The second type of chemicals of informative value avail-
able to aquatic prey are chemical alarm cues. These chemicals
are often released by prey animals when they are attacked or
captured by a predator (Chivers & Smith 1998). Such alarm
cues have been found in a wide variety of aquatic organisms,
both invertebrates (protozoans, flatworms, annelids, arthro-
pods, molluscs) and vertebrates (fishes and amphibians)
(reviewed by Wisenden 2003). They often elicit a dramatic
increase in antipredator behaviour when detected by con-
specifics and some heterospecifics (reviewed by Chivers &
Smith 1998). Experimental manipulations of alarm cue con-
centrations have shown that increased alarm cue concentra-
tions elicit increases in the intensity of antipredator behaviour
displayed by some fish (Dupuch et al. 2004, Zhao & Chivers
2006), including fathead minnows (Ferrari et al. 2005, Ferrari
& Chivers 2006). While these chemicals likely did not pri-
marily evolve as true alarm signals (Wisenden & Chivers
2005, Chivers et al. unpublished data), prey responding to
these chemicals have a selective advantage. Alarm cues
increase survival of receivers (Mathis & Smith 1993a; Mirza
& Chivers 2001). They also mediate learned predator recog-
nition through the pairing of alarm cues with novel predator
odours (Brown 2003).
The exact location of the alarm cues in the body of prey
is still unknown for most taxa. However, it has been shown
that fish chemical alarm cues are localized in their epidermis
(Chivers & Smith 1998). In addition to being released
Correspondence to: Maud. C. O. Ferrari, e-mail: maud.ferrari@usask.ca
373.qxd 7/24/2007 5:17 PM Page 157

158 M. C. O. Ferrari et al. CHEMOECOLOGY
through mechanical damage of the skin following a preda-
tory attack during which the fish is injured or captured
(capture-released), alarm cues have been shown to be
released in the feces of their predators (Mathis & Smith
1993b, 1993c, Brown et al. 1995) (diet-released). The alarm
cues or the breakdown products of the alarm cues are not
completely degraded during the digestion process.
While both capture- and diet-released alarm cues have
been shown to be important in mediating predator/prey
interactions, no studies have attempted to directly compare
the efficacy of the two cues in mediating antipredator
responses of the prey. Hence, the first goal of our experi-
ment was to compare the responses of fathead minnows to
capture- versus diet-released cues of predators in a con-
trolled experiment. The second goal of our experiment was
to investigate whether the type of predator (specialist or
generalist) would influence the intensity of these responses.
Here, we use two different fish predators, the northern
pike – a specialized piscivore, and brook trout (Salvelinus
fontinalis) – a more generalist predator. We hypothesized
that the characteristics of specialist piscivores (e.g., a large
mouth, pharyngeal suction etc.) would be associated with
shorter handling time of prey and would lead to a smaller
release of alarm cues during capture. Differential patterns of
cue release would likely influence the antipredator
responses of prey, as it has been shown that many fish
species increased the intensity of their antipredator response
when exposed to increased concentrations of conspecific
alarm cues (Ferrari et al. 2005). We hypothesized that if the
specialized piscivore (i.e., the pike) is effective at reducing
capture-released cues, then fathead minnows may effec-
tively consider pike as less of a threat than trout when the
opposite is true (Bertolo & Magnan 2006, Findlay et al.
2000). From the perspective of the prey, this may create a
paradox in terms of risk assessment, because fathead min-
nows assess local predation risk through alarm cue concentra-
tion. If specialized predators are more effective at breaking
down alarm cues through digestion than are generalist preda-
tors, then the prey should likewise consider the pike as less of
a threat. This could again create a paradox in terms of risk
assessment.
Methods
Fish collection and maintenance
Fathead minnows were collected from a one-ha pond located on
the University of Saskatchewan campus in Saskatoon, SK, Canada
in January of 2006 using Gee’s Improved Minnow Traps. The pond
contains fathead minnows and brook stickleback (Culaea incon-
stans) but no predatory fishes. Minnows from this pond do not
show innate recognition of fish predators including pike (Chivers
& Smith 1994). Immediately following capture minnows were
transported to the laboratory and maintained in a 518-L stream tank
at approximately 9
o
C and kept on a 15:9 h light:dark cycle. The
fathead minnows were fed a diet of commercial fish flakes daily.
The swordtails (Xiphophorus helleri) used in the experiment
came from multi-generational laboratory stock originally purchased
commercially from Florida, USA. The swordtails were maintained in
a 407-L tank at approximately 23
o
C and kept on a 14:10 h light:dark
cycle and fed a diet of commercial fish flakes daily.
The pike were collected using seine nets in the spring of 2005
from Pike Lake, SK., an oxbow lake of the South Saskatchewan River.
The pike were maintained in 75-L tanks containing approximately
37-L of water at 17
o
C and kept on a 15:9 h light:dark cycle for at least
two weeks prior to the experiment. In order to cleanse their digestive
tracts pike were initially denied food for a 10-day period and then
fed one convict cichlid (Archocentrus nigrofasciatus) (~3.5 cm fork
length) each every other day for a total of three feedings. We fed
convict cichlids to the predators as cichlids are not Ostariophysans
and do not possess alarm cues recognized by fathead minnows
(Chivers & Smith 1998, Brown et al. 2000). Moreover, cichlids are
distantly related to our control species (swordtails). Brook trout
(Salvelinus fontinalis) were obtained from the Fort Qu’Appelle
hatchery in September 2005 and were maintained and cleansed in
a fashion similar to the pike. Ten pike and ten trout were used to
prepare the stimuli.
The pike were approximately five cm longer than the trout
(~20 cm vs. ~15 cm standard length), however, the volumetric dis-
placement of the trout was about twice that of the pike: volumetric
displacement (mean ± S.D.) for pike = 74.6 ± 12.8 mL; trout =
157.8 ± 29.2 mL. Given that the two species of predators used in
this experiment are morphologically very different, it was difficult
to find a morphological characteristic we could use to match the
two species. Indeed, a match in predator body length would result
in comparing a fusiform slender pike to a deep-bodied trout.
Conversely, a match in predator volume would result in comparing
a short trout to a very long pike. None of these scenarios were sat-
isfying, thus we decided to compromise, using pike slightly longer
but more slender than trout.
Preparation of stimuli
To prepare the capture-released and diet-released cues, we used a
total of 10 pike and 10 trout. Half of the pike and half of the trout
were randomly assigned to be fed swordtail and the other half were
assigned to be fed minnows. A total of 30 swordtails and 30 min-
nows were used as prey; three prey items were fed to each of ten
trout and ten pike. Due to experimental constraints, we did not
measure the exact length of each prey fish. Indeed, manipulations
of live minnows that are not anaesthetized would elicit skin dam-
age and release of alarm cues. Alarm cells are easily damageable,
occurring on the outside of the fish scales. In addition, no anaes-
thetic could be used to measure the fish. A decrease in activity of
the test minnows due to potential contamination with anaesthetic in
the stimulus water could confound our results, as a decrease in
activity is a typical antipredator response in minnows. Thus, we
‘handpicked’ 30 swordtails and 30 minnows as similar in size as
possible (~5 cm standard length) and randomly assigned them as
prey for the two predator types.
The collection of stimuli was done in two phases. The objec-
tive of the first phase was to collect the capture-released cues. For
each prey type (i.e. swordtails and minnows), five pike and five
brook trout were removed from their holding tanks and placed into
individual aerated but not filtered 75-L tanks containing approxi-
mately 37 L of dechlorinated tap water. The pike and trout were fed
one prey (swordtail or minnow) at a time, each receiving three prey
fish in total. The handling time for each predation event was mea-
sured and recorded. Handling time was defined as the time of first
contact made by the pike or trout until oral manipulation of the
prey fish ceased and the mouth was closed. After each predator had
successfully consumed three prey, they were placed into a 9.4-L
tank filled with dechlorinated tap water for a five-minute period to
rinse any remnants of the capture-released cues from the body and
oral cavity of the predators. Immediately following this 5 min
period, the fish were moved to individual aerated but not filtered
75-L tanks (for the collection of diet-released cues) containing
approximately 37-L of water at 17
o
C. Following removal of the
predator from the capture-released collection tank, the water was
stirred vigorously and 500 mL from each of the 75-L tanks was
removed and placed in one of four communal receptacles (one for
swordtail cues from pike, one for minnow cues from pike, one
for swordtail cues from trout, one for minnow cues from trout).
The collected water was then stirred vigorously. From those stocks,
60-mL aliquots of stimulus were removed and frozen at -20
o
C.
The objective of the second phase was to collect the diet-
released cues. Stimulus was collected from each of the individual
tanks that the pike and trout were relocated to for three days after
collection of the capture-released cues. Diet-released cues were
373.qxd 7/24/2007 5:17 PM Page 158